Upload
phungdung
View
217
Download
2
Embed Size (px)
Citation preview
- 1 -
COMPARING ONLINE BEHAVIOR IN KOREA AND THE US YAHOO! ANSWERS VERSUS NAVER KNOWLEDGE-IN
Minhyung Kang*
Samsung Economic Research Institute, 1321-15, Seocho2-Dong, Seocho-Gu, Seoul, 137-955, Korea
Peter Gloor
Center for Collective Intelligence, Sloan School of Management, MIT, Cambridge, MA 02139, USA
Abstract
This study compares online behavior of the leading Q&A online sites in Korea and the
United States: Yahoo! Answers(US) and Naver Knowledge-iN (South Korea). We found
that Yahoo! Answers users communicate more actively than Naver Knowledge-iN users,
while Naver Knowledge-iN users seem more concerned about their social reputations.
Among three ties (i.e., “providing answers” tie, co-answering tie and “getting answers”
tie) at Q&A sites, “providing answers” ties show a positive influence on the user’s
quality of answers. Interestingly, co-answering ties show a negative influence on the
quality of answers. This implies that while an answerer’s successful communications with
askers are beneficial, relationships among co-answerers may be competitive. It also
seems that in online knowledge sharing, individualistic US users are more sociable than
collectivistic Korean users.
Keywords: Online social network; knowledge sharing; Q&A sites; cultural difference; web 2.0
* Corresponding author. Tel.: +822-3780-8225; Fax: +822-3780-1808; E-mail: [email protected]; Address: Samsung Economic Research Institute, Samsung Life Bldg. 31st Fl., 1321-15, Seocho2-Dong, Seocho-Gu, Seoul, 137-955, Korea
- 2 -
Introduction
Research on online social networks has been very active since the late 1990s (Ahuja & Carley, 1999;
Ahuja, Galletta, & Carley, 2003; Garton, Haythornthwaite, & Wellman, 1997; Thelwall, 2008; Wellman,
Haase, Witte, & Hampton, 2001; Wellman et al., 1996) with one of the main research topics being online
knowledge sharing (Subramani & Rajagopalan, 2003; Wasko & Faraj, 2000; Wasko & Faraj, 2005).
Wikipedia is a successful example of such an online knowledge sharing site (Stvilia, Twidale, Smith, &
Gasser, 2008; Wagner & Majchrzak, 2007). Wikipedia exists in many languages; the English language
Wikipedia for example boasts 2,772,824 articles (as of March 4, 2009) and 9,107,139 registered users.
The Wikipedia in Finnish, for instance, at the same time has 195 849 articles and 105 904 registered users.
The Korean Wikipedia, on the other hand, has 91,215 articles and 61,379 registered users. This is quite
amazing, because both countries are among the heaviest users of the Internet and both speak a language
that is unique and not shared by any other country. On the other hand, South Korea has 49 million
inhabitants, while Finland has 5 million. That means that Finland, a country with ten percent the
population of Korea, maintains a Wikipedia with twice the number of articles as Korea. Why is it that two
countries with comparable levels of technical sophistication and wealth show such vast differences in
Wikipedia creation and usage? We think that we can provide some answers to this puzzle by comparing
the knowledge acquisition and dissemination behavior between “Western” and ‘Eastern” cultures. In
particular, we hope to obtain some insights in online knowledge sharing behavior in Western countries
such as Finland and the US compared to an Asian country such as Korea.
As we are studying knowledge sharing activities, we are focusing on a different type of knowledge
sharing site with more process-oriented behavior, the “Question and Answer (Q&A) site.” While
Wikipedia provides general knowledge of interest to everyone, Q&A sites, such as Yahoo! Answers and
Naver Knowledge-In, provide customized knowledge to an individual asker. At the “Western” Yahoo!
Answers and “Eastern” Naver Knowledge-iN, millions of users are asking a question and providing
answers to others’ questions in various categories of topics (Adamic, Zhang, Bakshy, & Ackerman, 2008;
- 3 -
Harper, Raban, Rafaeli, & Konstan, 2008; Jurczyk & Agichtein, 2007). This is a unique benefit which
cannot be provided by Wikipedia. In addition, in a corporate environment, a successful Q&A site is an
efficient knowledge sharing platform for employees. Therefore, understanding facilitators of Q&A sites is
crucial to online knowledge sharing research. Especially, understanding drivers of core users’ knowledge
sharing at Q&A sites is critical because most of the contents at Q&A sites are provided by them. As it
turns out, the differences in online behavior of answerers and askers between Yahoo!Answers and Naver
are quite striking.
From a social network perspective, Q&A sites allow various relationships between askers and answerers
(e.g., answering relationship) as well as among answerers (e.g., co-answering relationship), while
Wikipedia provides relationships only among contributors (e.g., co-editing relationship). In other words,
Q&A sites provide more types of social relationships than Wikipedia. According to the theory of
embeddedness (Mark Granovetter, 1985), these social relationships may influence users’ answering
behaviors in various ways. By analyzing online social network data from two successful Q&A sites (i.e.,
Yahoo! Answers in the United States and Naver Knowledge-iN in South Korea), we would like to
validate online social network’ influence on the core users’ knowledge sharing behaviors at Q&A sites. In
particular, we also would like to explore differences in core users’ behavior at these Q&A sites embedded
in two different cultures.
To achieve these goals, we have the following research questions.
1. How do the properties of core users’ online social networks affect their performance at Q&A sites?
2. Are there any differences in core users’ behavior across different Q&A sites?
This research helps academics understand the motivating roles of social networks at Q&A sites,
especially in different cultural contexts. For practitioners, this paper can provide valuable guidelines for
launching new Q&A sites, or improve operations of existing Q&A sites.
- 4 -
This paper is organized as follows. First, we review the theoretical background of the research. Second,
the research model and hypotheses are introduced. Next, we explain the research methodology.
Afterward, the results of the data analysis are provided. Finally, we discuss the implications and
limitations of the research and suggest ideas for future research directions.
Theoretical Backgrounds and Hypotheses
Knowledge Sharing and Social Networks
Previous research on knowledge sharing from a social network perspective can be categorized into two
groups: The “strong-ties group” vs. the “weak-ties group”. Strength of ties represents the intensity of the
relationship (e.g., frequency of interaction, emotional closeness) between individuals or units (Rindfleisch
& Moorman, 2001; Brian Uzzi & Lancaster, 2003). The “strong-ties group” claims that strong ties are
likely to lead to active knowledge sharing among people (Hansen, 1999; Reagans & McEvily, 2003)
because it helps the formation of trust (D. Krackhardt, 1990), decreases the communication cost (B. Uzzi,
1997), and provides social motivations to be cooperative (M. S. Granovetter, 1973). On the other hand,
the “weak-ties group” focus more on the overall network properties of weak ties rather than the weakness
of ties. Weak ties need less effort to maintain and have more chances to occur than strong ties. Therefore,
weak ties allow a person to develop and keep relationships with many people (i.e., a large network) (Burt,
1983; Friedkin, 1982; Lin, Ensel, & Vaughn, 1981). Through this large network, people can access
various new ideas or resources that they may not get from their close friends. It means that the real value
of weak ties does not come from the weakness of ties, but from the size of the weak-tie network (i.e.,
overall structure of the network). The influence from the quality of relationships (e.g., strength of ties) is
referred as “relational embeddedness” and the influence from the structure of the overall social network
(e.g., size of network) is referred as “structural embeddedness (M. Granovetter, 1992).” Most studies
about knowledge sharing belong to the strong-ties group. On the other hand, studies about “online”
knowledge sharing emphasize weak ties (Table 1) because it is usually assumed that online relationships
are weak ties (Cummings, Butler, & Kraut, 2002). However, we think that online relationships can be
- 5 -
“strong” ties. For example, a core answerer in the inline skating category may have frequent interactions
with some askers and recognize them by their IDs. These relationships are surely different from the weak
ties the user has with other non-frequent users. Therefore, we consider the influence of both “strong”
online ties and “weak” online ties in this study.
[Insert Table 1 here]
Q&A sites: Yahoo! Answers and Naver Knowledge-iN
Q&A sites are online communities where users ask questions and answer others’ questions. In this paper,
we focus on two successful Q&A sites: Yahoo! Answers in the United States and Naver Knowledge-iN in
South Korea.
[Insert Table 2 here]
In these sites, an asker posts a question in a category and other users post answers for this question. When
the asker selects the best answer, this question is closed and moved to the resolved questions database. To
motivate a user’s participation, Q&A sites provide “knowledge points” to the user according to her
answering and asking performance. These points do not have any monetary value, but as in an online
game, the user’s level (or avatar (Shen, Radakrishnan, & Georganas, 2002)) will be upgraded as she gets
more points. In addition, a significant contributor’s ID is decorated with special icons such as “Hall of
Famer” or “Top Contributor” so that she will be recognized by other users whenever she posts.
Three Kinds of Ties
[Insert Figure 1 here]
From a user’s perspective, there are three kinds of ties at Q&A sites (Figure 1). First, a user can answer a
certain question asked by Asker A (i.e., answering tie from a focal user to Asker A). Second, another user
(Answerer B) can also answer the question asked by Asker A. In other words, the focal user and
Answerer B answer the same question. (i.e., co-answering tie between the focal user and Answerer B).
Lastly, the focal user can ask a question instead of answering. If this question is answered by Answerer C,
- 6 -
there is a “getting answers” tie from Answerer C to the focal user. Each of these three kinds of ties around
the focal user is expected to affect her knowledge sharing performance. To validate this claim, the
research model is composed of six research hypotheses about the average strength of ties and the size of
each of the three ties.
[Insert Figure 2 here]
Research hypotheses in Figure 2 are suggested based on the following arguments. First, “answering ties”
with many askers make the focal user familiar with the Q&A site’s group norms, feel integrated with the
site, and build up a sense of belonging to the site (O'Hara, Beehr, & Colarelli, 1994; Wasko & Faraj,
2005). This group identification makes the user care about the site’s success and motivates cooperative
behavior, such as knowledge sharing (Brewer & Kramer, 1986; Brewer & Schneider, 1990; Janssen &
Huang, 2007; Kramer & Brewer, 1984; Kramer & Goldman, 1995). On the other hand, repeated
“answering ties” with certain askers may make a user willing to spend some time and effort to help them
(Mark S. Granovetter, 1982; Reagans & McEvily, 2003). In addition, strong ties constitute a base of trust
(David Krackhardt, 1992; Levin & Cross, 2004; Tsai & Ghoshal, 1998), which gives the user the
confidence that the knowledge shared will not be misused (McEvily, Perrone, & Zaheer, 2003). The
willingness to help and trust in the askers can also lead to active knowledge sharing. Based on these
arguments, we have the following hypotheses:
Hypothesis 1: The more askers a user has answering ties with, the better the quality of his or her
answers will be.
Hypothesis 2: The more frequently a user interacts with askers, the better the quality of his or her
answers will be.
Second, if a user answers the same question jointly with other co-answerers, he will notice that other
users are also helping the asker and feel a kind of fellow feeling or homophily (McPherson, Smith-Lovin,
& Cook, 2001) with shared interests. This can be a positive motivation for knowledge sharing. Therefore,
- 7 -
if a user has many co-answerers, the quality of his answers will be better. In a similar vein, if a user
frequently interacts with a certain co-answerer, he may recognize the co-answerer and develop
interpersonal relationship using other communication channels, such as an e-mail. Through this personal
relationship, he can develop emotional closeness with the co-answerer and learn from the co-answerer.
Emotional attachment to other co-answerers is expected to strengthen the user’s positive attitude toward
knowledge sharing, and learning from the other co-answerers is expected to enhance the user’s expertise
to provide right answers. Therefore, the user’s average strength of co-answering ties may positively
influence the quality of shared knowledge.
On the contrary, co-answerers can be regarded as competitors because only one answer is selected as the
best answer among all the answers provided by answerers. Having many co-answerers means that there is
a little chance of being the best answerer. Moving to another category which does not have many
answerers could be a better choice for getting more knowledge points. In addition, if there are many co-
answerers, a user may feel less responsible for answering questions because there are many other users
who can answer (Chidambaram & Tung, 2005; Harkins & Szymanski, 1989). In a similar way, if a user
frequently comes across with certain co-answerers, she will recognize that there are other active or
competent answerers and may feel that her contribution is not necessary. Therefore, if she finds a question
answered by certain co-answerers, she may skip that question even though the question is not closed yet.
Based on these conflicting arguments, we have the following hypotheses. The latter two hypotheses are
alternative hypotheses for the former two expecting the opposite effects:
Hypothesis 3a: The more co-answerers a user has, the better the quality of his or her answers
will be.
Hypothesis 4a: The more frequently a user interacts with co-answerers, the better the quality of
his or her answers will be.
Hypothesis 3b: The fewer co-answerers a user has, the better the quality of his or her answers
will be.
- 8 -
Hypothesis 4b: The less frequently a user interacts with co-answerers, the better the quality of his
or her answers will be.
Last, if a user gets answers for his questions at a Q&A site, he may realize that the site is useful and start
to contribute (i.e., share knowledge) to keep this site alive (Lawler & Yoon, 1996). Also, once a user gets
some help from other users, he may feel responsible to reciprocate by contributing his own knowledge
(Gouldner, 1960). Besides, the knowledge from others can improve the user’s expertise so that the quality
of the user’s answers will be enhanced. To sum up, many helpful relationships with other users can
motivate a user to actively contribute his knowledge in return. In addition, if a user gets help from the
same answerers frequently, he can recognize those answerers. This recognition of benevolent other users
can lead to the user’s emotional attachment to them and motivate the user to reciprocate their help with
his contribution. Based on these arguments, we have the following hypotheses:
Hypothesis 5: The more answerers a user has for his or her questions, the better the quality of his
or her answers will be.
Hypothesis 6: The more frequently a user gets answers from other users, the better the quality of
his or her answers will be.
Research Methodology
The sample consists of the top 100 contributors each of Yahoo! Answers and Naver Knowledge-iN. Some
of Yahoo! Answers users hide their questions and answers list, so they are excluded from the sample and
the next top contributors are included. Therefore, the total sample size is 200 (i.e., 100 for each site).
Using a web crawler, we collected questions and answers of each of 200 contributors. At first, we
analyzed the descriptive properties of these datasets and compared them across Yahoo! Answers and
Naver Knowledge-iN. Then we conducted a regression analysis to verify our research hypotheses.
- 9 -
Measurement
For every posting, we gathered information about title, asker, answerer, timestamp, selection (i.e.,
selected as the best answer or not), and content. Based on these online data, all measures in this paper are
objectively calculated.
Network Measures
Network measures are calculated from the three kinds of ties at Q&A sites. First, an “answering” tie
exists between a focal user and an asker when the asker chooses the focal user’s answer as the best
answer. This is a direct relationship between the answerer and the asker. Second, a “co-answering” tie
exists between a focal user and another user when both of them answer the same question. This is an
indirect relationship between users by answering the same question. Last, a “getting answers” tie exists
between a focal user and an answerer when the focal user chooses the answerer’s answer as the best
answer. This is another direct relationship between the answerer and the asker, but the direction is the
opposite of an “answering” tie. For each of these three ties, we calculate structural and relational
embeddedness using degree centrality (Freeman, 1979) and average strength of ties (Marsden &
Campbell, 1984). Degree centrality in an answering network is the total number of askers the focal user
has successfully helped (i.e., provided the best answers). If a user posted 30 best answers to the questions
of 5 askers, degree centrality in an answering network is 5 because the user built the relationship with
those 5 askers. So, degree centrality shows how many askers a focal user has successful relationships with.
However, it does not show how frequent (or strong) the relationship between the focal user and the askers
is. Average strength of ties covers this. In the previous example, the user posted 30 best answers to the
questions of 5 askers. In other words, in average, the focal user had 6 successful interactions with each
asker. This is a quite high number and the focal user and the askers may recognize each other. This
“strong” tie may have unique influence on the focal user’s knowledge sharing performance. In the same
way, degree centrality and average strength of ties are calculated for the other two ties (i.e., “co-
answering” ties, “getting answers” ties).
- 10 -
Dependent Variable
The quality of the answers is calculated by an individual user’s rate of best answers over total answers.
For example, if a user has 50 best answers out of 100 answers, the quality of her answers is 0.5. Even
though another user has the same number of best answers (i.e., 50 best answers), if his total number of
answers is 1000, the quality of his answers is only 0.05, which is far beyond the former user’s quality of
answers. To verify a causal relationship between network measures and the quality of answers, network
measures are calculated for the earlier half of a user’s lifecycle (i.e., from the first posting time to (the
first posting time + the last posting time)/2) and quality of the answers is calculated for the later half of
the user’s lifecycle.
Control Variable
We controlled the influence of social reputation. When a user is publicly awarded or recognized as a
“Hall of Famer” or “Top Contributor,” she may feel more responsibility to contribute and put more effort
into her answers. This can influence the answering performance of users regardless of the properties of
the users’ social networks. Therefore, we added social reputation as a control variable to focus on the
effects of social networks.
Results
Descriptive Findings and Differences across Yahoo! Answers and Naver Knowledge-iN
Through January 2008, 1,826,075 questions and 24,517,045 answers had been gathered at Yahoo!
Answers and 868,887 questions and 1,736,218 answers at Naver Knowledge-iN. Considering the gap
between the launching dates of the two sites (October 2002 and December 2005), Yahoo! Answers is
expanding very fast. At first, we discuss common characteristics of the two Q&A sites. Then differences
between the two sites are explored.
First, at both sites, top contributors are biased toward answering and spend little time asking. In other
words, top contributors are not top askers. The ratios of questions to answers per user are 0.53% at
- 11 -
Yahoo! Answers and 0.40% at Naver Knowledge-iN (Table 3). Therefore, top contributors participate in
Q&A sites to help others, not to get help from others.
[Insert Table 3 here]
Second, questioning and answering relationships among the top contributors are very rare at both sites.
This is mainly because there are few questions from the top contributors. In addition, top contributors do
not seem to differentiate between other top contributors’ answers and normal users’ answers. Among the
top contributors of Yahoo! Answers, for 7,933 questions, 20,469 answers were given and 1,358 answers
were chosen as the best answers (The rate of selection: 6.6%). Among the top contributors of Naver
Knowledge-iN, for 2,824 questions, 2,086 answers were given and 1,716 answers were chosen as the best
answers (The rate of selection: 82.3%). Therefore, the rates of selection among top contributors are
similar to the rates of selection by normal users (i.e., 7.8% at Yahoo! Answers and 83.3% at Naver
Knowledge-iN). They may communicate by other communication channels such as e-mail or online
memo, but they do not show many direct questioning and answering relationships among themselves.
However, there are considerable differences between Yahoo! Answers and Naver Knowledge-iN and all
the differences are proved significant by one-way ANOVA test (Table 4). The average rates of selection
are 7.8% at Yahoo! Answers and 83.3% at Naver Knowledge-iN. It seems that top contributors of Yahoo!
Answers are not much concerned about the rate of selection, while top contributors of Naver Knowledge-
iN are much concerned about it. At Yahoo! Answers, users answer any questions they have opinions on,
even though their answers might overlap with other answers or not provide unique contributions.
However, at Naver Knowledge-iN, if questions are correctly answered once, other users rarely post more
answers to those questions. The average number of answers per question (12.43 at Yahoo! Answers and
2.00 at Naver Knowledge-iN) also shows that users of Yahoo! Answers answer more frequently than
users of Naver Knowledge-iN. Instead, the average length of answers is 190 bytes at Yahoo! Answers and
799 bytes at Naver Knowledge-iN. For instance, users of Yahoo! Answers usually describe their opinion
in a couple of sentences off the top of their heads, but users of Naver Knowledge-iN frequently search
- 12 -
and refer to other external sources to complement their answers in order to be more accurate. In other
words, it seems that top contributors of Yahoo! Answers answer more frequently, while top contributors
of Naver Knowledge-iN put more effort into answering each question.
[Insert Table 4 here]
For each site, correlations among the variables are explored (Table 5). We checked for multicollinearity
because there are relatively high correlations among some of the variables (e.g., centrality of answering
ties and centrality of co-answering ties at Naver, centrality of answering ties and quality of answers at
Yahoo). The variance inflation factor (VIF) values for all of the variables are less than 2.04, which shows
that multicollinearity is not a serious issue in our study
[Insert Table 5 here]
Regression Results and Hypothesis Testing
Table 6 reports the effects of the properties of online networks on the quality of answers at each Q&A site.
[Insert Table 6 here]
First, the centrality of answering ties significantly influences the quality of answers at both Yahoo!
Answers and Naver Knowledge-iN. However, the average strength of answering ties does not have a
significant effect at either site. Therefore, for answering ties, structural embeddedness is supported
(Hypothesis 1), but relational embeddedness is not supported (Hypothesis 2).
Second, the centrality of co-answering ties negatively influences the quality of answers at both sites. The
effect of the average strength of co-answering ties is also negatively significant, but only at Naver
Knowledge-iN. Therefore, co-answering ties’ structural embeddedness is supported (Hypothesis 3b) and
relational embeddedness is partially supported (Hypothesis 4b). Contrary to the effect of answering ties,
the effect of co-answering ties on the quality of answers seems negative.
- 13 -
Last, getting answers ties show no significant influence on the quality of knowledge at either site.
Therefore, no hypothesis for getting answers ties is supported (Hypothesis 5 and 6). Insignificant effect of
getting answers ties implies that top contributors of Q&A sites help other users because of enjoyment in
helping, social reputation, etc., and not because of willingness to reciprocate.
Limitations and Directions for Future Research
We acknowledge that this study has some limitations. First, we could not include some of the top 100
contributors of Yahoo! Answers within the sample because they have hidden their Q&A list. (This
functionality is available only at Yahoo! Answers.) We had to replace them with the next top contributors
to have the same number of contributors at both sites. Additionally, we could not gather all the answers of
some of the top contributors because their numbers of answers exceeded the maximum number of
postings accessible by general users (i.e., 30,000). We excluded those seven contributors from the sample
and conducted a regression analysis again, but the result was similar. Consequently, we report the original
result with 100 contributors for each site. Second, our data analysis includes questioning and answering
relationships among users only and does not include other interactions using private communication
channels such as e-mail or online memo. Even though they seem to communicate with each other through
these private communication channels, we could not utilize those data due to privacy concerns. With the
users’ agreements and Q&A sites’ cooperation, they may be analyzed together in the future. Third, since
our samples were drawn only from the top contributors of two Q&A sites, the results of the study may not
be generalizable to normal contributors or sites of different nationalities. To overcome this limitation,
comparing the top contributors’ group and the normal (or infrequent) contributors’ group would be an
interesting topic for future research. Lastly, this research did not cover a comprehensive list of potential
factors that may influence the quality of answers at Q&A sites because it focused on online social
networks among users. For instance, latent variables such as group identification or self-efficacy (Bock,
Zmud, Kim, & Lee, 2005; Kankanhalli et al., 2005) may also influence a contributor’s quality of answers.
- 14 -
An online survey to capture the top contributors’ subjective data could be considered to extend the current
study.
Discussion
In this paper, we explored the top contributors’ behaviors of two successful Q&A sites, Yahoo!
Answers and Naver Knowledge-iN, and compared the differences in their behaviors between the
two sites. Using a regression analysis, we also verified the effects of online social networks on the
quality of answers at the two sites.
Users on both the US and the Korean Q&A site exhibit some common characteristics. Online Q&A
data from Yahoo! Answers and Naver Knowledge-iN shows that the users at Q&A sites are not
homogeneous. There are some core contributors who focus on answering other users’ questions.
Their voluntary behaviors are not based on expected reciprocity because they rarely ask questions.
They just enjoy helping others or being recognized as experts by other users (Constant, Sproull, &
Kiesler, 1996; Kankanhalli, Tan, & Kwok-Kee, 2005; Wasko & Faraj, 2005). To identify and
motivate these voluntary contributors is critical at Q&A sites because they create most of the
contents. However, askers also matter because they trigger answerers’ contributions by posting
questions and appreciate their contributions by choosing the best answers. So, keeping the proper
number of active askers and guiding them to ask proper questions and to choose right answers as
the best answers are also imperative. Practitioners may need separate strategies to take care of these
two groups individually.
In addition, the main relationships at Q&A sites are questioning and answering relationships
between askers and answerers. There are few direct interactions among top contributors. This is a
unique characteristic of Q&A sites. In other online or offline communities, core members
frequently interact with each other, as well as interacting with other peripheral members (Borgatti &
Everett, 2000). However, at Q&A sites, top contributors have many direct interactions with askers,
but few direct interactions among themselves. This may indicate competitive relationships among
co-answerers; they rarely interact with each other and perhaps are competing in order to become the
best answerer. Considering the successful collaborations among contributors at Wikipedia, Q&A
sites should facilitate positive interactions among contributors. For example, Q&A sites could hold
offline gatherings of top contributors, like Wikimania
(http://wikimania2008.wikimedia.org/wiki/Main_Page), or provide online chat rooms to support
real-time communications among contributors while they are posting their answers.
The two Q&A sites also show marked differences in top contributors’ behaviors. Western top
contributors of Yahoo! Answers ask and answer questions more frequently, while Korean top
contributors of Naver Knowledge-iN seem to be more concerned about being the best answerers
and put more effort into answering each question. Cultural differences (Li, Li, & Lin, 2008)
between the United States and Korea may be one of the plausible reasons. Korean people have a
more collective culture than the US people (G. Hofstede, 1984), and they are more concerned about
social reputation or losing face in public (Kim & Nam, 1998). This may also apply to online sites.
Thus, Korean people may put more effort in answering each question so that they do not lose face
by providing faulty or low-quality answers. To the contrary, the US people seem to be more
comfortable expressing and discussing their different opinions in public (Barker, 1997; G. J.
Hofstede, 2005).
Another plausible reason is the existence of alternative knowledge sharing sites. In the United
States, Wikipedia has been very popular since its launch in 2001, while, as has been shown in the
introduction, Wikipedia is not very popular in Korea. Serious answerers in the United States may
therefore contribute to Wikipedia instead of Yahoo! Answers because Yahoo! Answers seems more
discussion-oriented (Adamic et al., 2008). On the other hand, in Korea, Naver Knowledge-iN is the
dominant site for any kind of online knowledge sharing. This might again reflect a deep cultural
difference, because an individual answerer in Naver can build up a reputation as a “hall of famer”,
while contributing to Wikipedia does not build up an individual’s reputation. Rather, Wikipedia
recognizes “Today’s featured article”, which very much recognizes the achievements of a group.
This is in contradiction to general wisdom that would assign a collectivist culture to Eastern
countries like Korea, while individualistic behavior is more expected in Western countries such as
the US (G. Hofstede, 1984). It seems, however, that in knowledge sharing processes this behavior
might be reverted between East and West.
Lastly, different user policies are also worth mentioning. While Naver Knowledge-iN limits the
maximum number of postings for any users (e.g., the maximum number of answers per day is 10,
30, or 50 according to the user’s knowledge level) (Naver), Yahoo! Answers puts no limit if a user
reaches a certain level (e.g., level 5) (Answers). This may make users of Naver Knowledge-iN put
more effort into each answer. To improve the quality of answers, Yahoo! Answers may consider
limiting the maximum number of postings even for the high-level users.
Regression analysis of Q&A data also provides interesting implications to researchers by verifying
the different effects of the three types of relationships at Q&A sites. While the number of answering
ties with askers (i.e., degree centrality) shows a positive influence on the quality of answers,
average strength of answering ties does not show a significant effect. It implies that what matters at
Q&A sites for a core contributor is interactions with general askers, not personal interactions with
individual askers. Successful interactions with these general askers may lead to users’ commitment
to overall Q&A sites.
On the other hands, co-answering relationships with other answerers show a negative influence on
the quality of answers. These indirect relationships among answerers do not seem to build a group
solidarity or fellow feeling. Instead, these relationships make users realize that their chances of
being selected as the best answerers are low and also make users feel less responsible or willing to
answer questions because they know that there are other answerers who can answer these questions.
This is a unique finding of this study because these negative roles of social relationships are rarely
explored in the knowledge management literature. Lastly, getting answers ties do not significantly
influence the quality of answers. Compared with the large volume of answering ties, relatively
small numbers of getting answers ties do not seem to have any impact on the quality of answers.
This indicates that top contributors of Q&A sites answer questions because they enjoy helping
others or being recognized by other users, not because they feel responsible to reciprocate other
users’ previous help. They may regard Q&A sites as the places to help other users, not the place to
exchange knowledge among each other.
Besides the theoretical implications, these findings also provide practical guidelines to improve the
quality of answers at Q&A sites. To strengthen the positive influence of answering ties, Q&A sites
may notify users of new answers from the previous best answerers for their questions, or new
questions from the previous askers who chose the users’ answers as the best answers. This may
lead to more active relationships between askers and answerers. Yahoo! Answers recently added a
similar functionality and we believe that this is the right choice for the site. Additionally, Q&A sites
should try to convert the negative effect of co-answering relationships into a positive one by
changing competitive relationships among answerers to collaborative or cooperative relationships.
As previously described, providing more chances for online or offline gatherings of answerers will
be beneficial to the quality of answers because these gatherings will facilitate knowledge sharing
among top contributors. Q&A sites may also consider a collaborative answering environment,
which new Q&A sites, such as WikiAnswers (http://wiki.answers.com/), are trying to provide to
integrate advantages of both Wikipedia and Q&A sites.
Acknowledgement
The authors would like to thank Tom Malone and Rob Laubacher for their feedback and for
providing a stimulating environment at MIT Center for Collective Intelligence. We are also grateful
to Lada Adamic for her advice on collecting data from Yahoo! Answers.
References
Adamic, L.A., Zhang, J., Bakshy, E., & Ackerman, M.S. (2008). In Knowledge sharing and yahoo
answers: Everyone knows something. Paper presented at the WWW 2008, Beijing, China. ACM.
Ahuja, M.K., & Carley, K.M. (1999). Network structure in virtual organizations. Organization
Science, 10(6), 741-757.
Ahuja, M.K., Galletta, D.F., & Carley, K.M. (2003). Individual centrality and performance in
virtual r&d groups: An empirical study. Management Science, 49(1), 21-38.
Answers, Y. Points and levels. Retrieved June 11, 2008, from
http://answers.yahoo.com/info/scoring_system;_ylt=Aj50sRnE_dmUwjEB6i8ERKT59xd.
Barker, J. (1997). The purpose of study, attitudes to study and staff-student relationships. In D.
McNamara & R. Harris (Eds.), Overseas students in he: Issues in teaching and learning. London:
Routledge.
Bock, G.-W., Zmud, R.W., Kim, Y.-G., & Lee, J.-N. (2005). Behavioral intention formation in
knowledge sharing: Examining the roles of extrinsic motivators, social-psychological forces, and
organizational climate. MIS Quarterly, 29(1), 87-111.
Borgatti, S.P., & Everett, M.G. (2000). Models of core/periphery structures. Social Networks,
21(4), 375-395.
Brewer, M.B., & Kramer, R.M. (1986). Choice behavior in social dilemmas: Effect of social
identity, group size, and decision framing. Journal of Personality and Social Psychology, 50, 543-
540.
Brewer, M.B., & Schneider, S. (1990). Social identity and social dilemmas: A double-edged sword.
In D.Abrams & M.A. Hogg (Eds.), Social identity theory: Constructive and critical advances (pp.
22-41). New York: Springer-Verlag.
Burt, R.S. (1983). Range. In R.S. Burt & M.J. Minor (Eds.), Applied network analysis: A
methodological introduction (pp. 176-194). Beverly Hills: Sage Publications.
Chidambaram, L., & Tung, L.L. (2005). Is out of sight, out of mind? An empirical study of social
loafing in technology-supported groups. Information Systems Research, 16(2), 149-168.
Constant, D., Sproull, L., & Kiesler, S. (1996). The kindness of strangers: The usefulness of
electronic weak ties for technical advice. Organization Science, 7(2), 119-135.
Cummings, J.N., Butler, B., & Kraut, R. (2002). The quality of online social relationships.
Communications of the ACM, 45(7), 103-108.
Freeman, L.C. (1979). Centrality in social networks: Conceptual clarification. Social Networks, 1,
215-239.
Friedkin, N. (1982). Information flow through strong and weak ties in intraorganizational social
networks. Social Networks, 3(4), 273-285.
Garton, L., Haythornthwaite, C., & Wellman, B. (1997). Studying online social networks. Journal
of Computer-Mediated Communication, 3(1), Online.
Gouldner, A.W. (1960). The norm of reciprocity: A preliminary statement. American Sociological
Review, 25(2), 161-178.
Granovetter, M. (1985). Economic action and social structure: The problem of embeddedness. The
American Journal of Sociology, 91(3), 481-510.
Granovetter, M. (1992). Problems of explanation in economic sociology. In N. Nohria & R.G.
Eccles (Eds.), Networks and organizations: Structure, form, and action (pp. 25-56). Boston,
Massachusetts: Harvard Business School Press.
Granovetter, M.S. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360.
Granovetter, M.S. (1982). The srength of weak ties: A network theory revisited. In P.V. Marsden &
N. Lin (Eds.), Social structure and network analysis. Beverly Hills: Sage Publications.
Hansen, M.T. (1999). The search-transfer problem: The role of weak ties in sharing knowledge
across organization subunits. Administrative Science Quarterly, 44(1), 82-111.
Harkins, S.G., & Szymanski, K. (1989). Social loafing and group evaluation. Journal of personality
and social psychology, 56(6), 934-941.
Harper, F.M., Raban, D., Rafaeli, S., & Konstan, J.A. (2008). In Predictors of answer quality in
online q&a sites (pp. 865-874). Paper presented at the The Twenty-Sixth Annual SIGCHI
Conference on Human Factors in Computing Systems, Florence, Italy. ACM New York, NY, USA.
Hofstede, G. (1984). Culture's consequences: International differences in work-related values.
Newbury Park: Sage Publications Inc.
Hofstede, G.J. (2005). Cultures and organizations: Software for the mind. New York: McGraw-Hill.
Janssen, O., & Huang, X. (2007). Us and me: Team identification and individual differentiation as
complementary drivers of team members' citizenship and creative behaviors. Journal of
Management, 33(5), 724-751.
Jurczyk, P., & Agichtein, E. (2007, July 23–27). In Hits on question answer portals: Exploration of
link analysis for author ranking (pp. 845-846). Paper presented at the ACM SIGIR, Amsterdam,
The Netherlands. ACM Press New York, NY, USA.
Kankanhalli, A., Tan, B.C.Y., & Kwok-Kee, W. (2005). Contributing knowledge to electronic
knowledge repositories: An empirical investigation. MIS Quarterly, 29(1), 113-143.
Kim, J.Y., & Nam, S.H. (1998). The concept and dynamics of face: Implications for organizational
behavior in asia. Organization Science, 9(4), 522-534.
Krackhardt, D. (1990). Assessing the political landscape: Structure, cognition, and power in
organizations. Administrative Science Quarterly, 35(2), 342-369.
Krackhardt, D. (1992). The strength of strong ties: The importance of philos in organizations. In N.
Nohria & R. Eccles (Eds.), Networks and organizations (pp. 216-239). Boston: Harvard Business
School Press.
Kramer, R.M., & Brewer, M.B. (1984). Social group identity on resource use in a simulated
commons dilemmas. Journal of Personality and Social Psychology, 46(5), 1044-1057.
Kramer, R.M., & Goldman, L. (1995). Helping the group or helping yourself? In D. Schroeder
(Ed.), Social dilemmas (pp. 49-68). New York: Praeger.
Lawler, E.J., & Yoon, J. (1996). Commitment in exchange relations: Test of a theory of relational
cohesion. American Sociological Review, 61(1), 89-108.
Levin, D.Z., & Cross, R. (2004). The strength of weak ties you can trust: The mediating role of trust
in effective knowledge transfer. Management Science, 50(11), 1477-1490.
Li, D., Li, J., & Lin, Z. (2008). Online consumer-to-consumer market in china - a comparative study
of taobao and ebay. Electronic Commerce Research and Applications, 7(1), 55-67.
Lin, N., Ensel, W.M., & Vaughn, J.C. (1981). Social resources and strength of ties: Structural
factors in occupational status attainment. American Sociological Review, 46(4), 393-405.
Marsden, P.V., & Campbell, K.E. (1984). Measuring tie strength. Social Forces, 63(2), 482-501.
McEvily, B., Perrone, V., & Zaheer, A. (2003). Trust as an organizing principle. Organization
Science, 14(1), 91-103.
McPherson, M., Smith-Lovin, L., & Cook, J.M. (2001). Birds of a feather: Homophily in social
networks. Annual Review of Sociology, 27(1), 415-444.
Naver. (June 10, 2008). How many points can i get when i use naver knowledge-in? Retrieved June
11, 2008, from http://help.naver.com/service/svc_index.jsp?selected_nodeId=NODE0000000071
O'hara, K.B., Beehr, T.A., & Colarelli, S.M. (1994). Organizational centrality: A third dimension of
intraorganizational career movement. Journal of Applied Behavioral Science, 30(2), 198-216.
Reagans, R., & McEvily, B. (2003). Network structure and knowledge transfer: The effects of
cohesion and range. Administrative Science Quarterly, 48(2), 240-267.
Rindfleisch, A., & Moorman, C. (2001). The acquisition and utilization of information in new
product alliances: A strength-of-ties perspective. Journal of Marketing, 65(2), 1-18.
Shen, X., Radakrishnan, T., & Georganas, N.D. (2002). Vcom: Electronic commerce in a
collaborative virtual world. Electronic Commerce Research and Applications, 1(3-4), 281-300.
Stvilia, B., Twidale, M.B., Smith, L.C., & Gasser, L. (2008). Information quality work organization
in wikipedia. Journal of American Society for Information Science and Technology, 59(6), 983.
Subramani, M.R., & Rajagopalan, B. (2003). Knowledge-sharing and influence in online social
networks via viral marketing. Communications of the ACM, 46(12), 300-307.
Thelwall, M. (2008). Social networks, gender, and friending: An analysis of myspace member
profiles. Journal of American Society for Information Science and Technology, 59(8), 1321.
Tsai, W., & Ghoshal, S. (1998). Social capital and value creation: The role of intrafirm networks.
Academy of Management Journal, 41(4), 464-476.
Uzzi, B. (1997). Social structure and competition in inter-firm networks: The paradox of
embeddedness. Administrative Science Quarterly, 42.
Uzzi, B., & Lancaster, R. (2003). Relational embeddedness and learning: The case of bank loan
managers and their clients. Management Science, 49(4), 383-299.
Wagner, C., & Majchrzak, A. (2007). Enabling customer-centricity using wikis and the wiki way.
Journal of Management Information Systems, 23(3), 17-43.
Wasko, M.M., & Faraj, S. (2000). "It is what one does": Why people participate and help others in
electronic communities of practice. The Journal of Strategic Information Systems, 9(2-3), 155-173.
Wasko, M.M., & Faraj, S. (2005). Why should i share? Examining social capital and knowledge
contribution in electronic networks of practice. MIS Quarterly, 29(1), 35-57.
Wellman, B., Haase, A.Q., Witte, J., & Hampton, K. (2001). Does the internet increase, decrease, or
supplement social capital?: Social networks, participation, and community commitment. American
Behavioral Scientist, 45(3), 436-455.
Wellman, B., Salaff, J., Dimitrova, D., Garton, L., Gulia, M., & Haythornthwaite, C. (1996).
Computer networks as social networks: Collaborative work, telework, and virtual community.
Annual Reviews in Sociology, 22(1), 213-238.
- 21 -
Tables and Figures
Table 1
Yahoo! Answers and Naver Knowledge-iN
Yahoo! Answers Naver Knowledge-iN
URL http://answers.yahoo.com http://kin.naver.com
Launching Date
December, 2005 October, 2002
Market Share in host country
74.05% (as of March 19, 2008, by Hitwise.com) [40]
86.02% (as of May 07, 2008, by Rankey.com)
Table 2
Difference between Q&A sites and Wikipedia
Properties of Knowledge Shared
Properties of Social Networks
Q&A sites Individual Knowledge
(+) Customized Knowledge
(-) Redundant Knowledge
Nodes: answerer, asker
Ties: ask, answer, co-answer, vote, object, etc.
Wikipedia Collective Knowledge
(+) Quality
(+) Format/Style
Nodes: contributor
Ties: co-edit, talk (article/personal)
Table 3
Q&A Sites’ Statistics of Activities (1)
Avg. # of Answers per User
Avg. # of Questions per User
Avg. Ratio of Q/A per
User
Yahoo 14,838 79 0.53%
Naver 7,011 28 0.40%
- 22 -
Table 4
Q&A Sites’ Statistics of Activities (2)
Avg. Rate of
Selection
Avg. Length of Answers
Avg. # of Answers per
Question
Yahoo 7.8% 190 (bytes) 12.43
Naver 83.3% 799 (bytes) 2.00
Table 5
Correlations at Q&A sites
(Lower Diagonal: Yahoo! Answers, Upper Diagonal: Naver Knowledge-iN)
1 2 3 4 5 6 7
1. Centrality of Answering Ties 0.577** 0.259** 0.102 -0.082 0.097 0.040
2. Centrality of Co-Answering Ties 0.124 0.156 -0.323** -0.083 0.237* -0.451**
3. Centrality of Getting Answers Ties 0.068 0.356** -0.056 -0.065 0.254* -0.099
4. Avg. Strength of Answering Ties 0.451** -0.210* -0.064 0.136 -0.098 0.053
5. Avg. Strength of Co-Answering Ties 0.047 0.248* 0.287** 0.157 -0.209* 0.053
6. Avg. Strength of Getting Answers Ties 0.149 0.184 0.478** -0.032 0.260** -0.129
7. Quality of Answers 0.517** -0.407** -0.073 0.343** -0.005 0.058
- 23 -
Table 6
Regression of the Quality of Answers on Network Measures at Q&A sites
(1. Coefficients are standardized. 2. **: p < 0.01, *: p < 0.05)
Yahoo! Answers Naver Knowledge-iN
Variables Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Social Reputation .357 .113 .117 .309 .235 .233
Centrality of Answering Ties
.530** .559** .450 .601**
Centrality of Co-Answering Ties
-.490** -.527** -.675 -.866**
Centrality of Getting Answers Ties
.085 .059 -.092 -.131
Avg. Strength of Co-Answering Ties
.113 .066
Avg. Strength of Answering Ties
-.068 -.309**
Avg. Strength of Getting Answers Ties
.001 .065
Adj. R2 .119 .496 .492 .095 .404 .478
Sig. F Change .000 .000 .531 .002 .000 .007
# of cases 100 100 100 100 100 100
Fig. 1. Three Kinds of Ties
Fig. 2. Research Model
1. Answering Tie with A
2. Co-Answering Tie with B
Focal User
Answer
Asker A
Get an Answer
Answerer C
Answer
Answerer B
3. Getting Answers Tie with C
Co-
Answ
er
1. Answering Tie with A
2. Co-Answering Tie with B
Focal User
Answer
Asker AAsker A
Get an Answer
Answerer CAnswerer C
Answer
Answerer BAnswerer B
3. Getting Answers Tie with C
Co-
Answ
er
Quality of Answers
H1
H4
Centrality of Answering Ties
Centrality of Co-Answering Ties
Avg. Strength ofAnswering Ties
Avg. Strength ofCo-Answering Ties
H3
H5
H2
Centrality of Getting Answers Ties
Avg. Strength ofGetting Answers Ties
Dependent Variable
Structural
Embeddedness
Relational
Em
beddedness
H6
Quality of Answers
H1
H4
Centrality of Answering Ties
Centrality of Co-Answering Ties
Avg. Strength ofAnswering Ties
Avg. Strength ofCo-Answering Ties
H3
H5
H2
Centrality of Getting Answers Ties
Avg. Strength ofGetting Answers Ties
Dependent Variable
Structural
Embeddedness
Relational
Em
beddedness
H6